Chainer
v3.2.0
Tutorial
Introduction to Chainer
How to Write a New Network
How to Train a Network
How to write a training loop in Chainer
Let’s try using the Trainer feature
Using GPU(s) in Chainer
Customizing Chainer
Type check
Reference Manual
Development
API Compatibility Policy
Contribution Guide
Misc
Installation Guide
Tips and FAQs
Upgrade Guide from v1 to v2
Comparison with Other Frameworks
License
Chainer
Docs
»
Tutorial
»
How to Train a Network
Edit on GitHub
How to Train a Network
¶
How to write a training loop in Chainer
1. Prepare a dataset
2. Create a dataset iterator
3. Define a network
4. Select an optimization algorithm
5. Write a training loop
6. Save the trained model
7. Perform classification by the saved model
Let’s try using the Trainer feature
1. Prepare the dataset
2. Prepare the dataset iterations
3. Prepare the model
4. Prepare the Updater
5. Setup Trainer
6. Add Extensions to the Trainer object
7. Start Training
8. Evaluate a pre-trained model